NVIDIA Agrees to Acquire Hugging Face for $12.93 Billion

NVIDIA has entered into a definitive agreement to acquire Hugging Face in a transaction announced at $12,930,300,000, bringing one of the AI industry’s largest repositories of models, datasets and applications under the control of its dominant accelerator supplier.

The agreement was signed on September 2, 2026, according to NVIDIA’s Form 8-K filing with the US Securities and Exchange Commission. The companies announced it publicly the following day. They expect the transaction to close in the first half of 2027, subject to regulatory approvals and other customary closing conditions.

The headline value does not consist entirely of proceeds for Hugging Face’s owners. NVIDIA’s filing divides the transaction into an approximately $11.9 billion purchase price for Hugging Face stockholders, subject to adjustments, and an equity-based retention program worth up to approximately $1 billion for employees who join NVIDIA.

Hugging Face co-founder and CEO Clément Delangue said the founders and team intend to remain with the business. He presented the acquisition as a way to obtain more computing capacity, support and visibility for open-source AI while keeping the platform open and compute-agnostic.

1. What NVIDIA is buying

Founded in 2016 by Delangue, Julien Chaumond and Thomas Wolf, Hugging Face has developed into a central distribution and collaboration platform for machine-learning software. Its Hub hosts models, datasets and interactive applications, while its commercial products support model evaluation, customization and deployment.

NVIDIA says more than 18 million developers, researchers and creators use Hugging Face. The platform contains more than 3 million models, 500,000 datasets and 1 million applications, according to figures published with the acquisition announcement. More than 200,000 companies use it to discover, evaluate, customize or deploy AI.

Those figures explain why the transaction is about more than purchasing a software vendor. Hugging Face sits between model creators and the organizations that download, modify and deploy their work. It provides infrastructure through which developers encounter new architectures, libraries, datasets and deployment options.

The proposed purchase follows years of investment in the company. Axios reports that Hugging Face had raised approximately $400 million and was most recently valued at $4.5 billion in 2023. The approximately $11.9 billion earmarked for stockholders is roughly 2.6 times that valuation, although the figures are not directly equivalent because the acquisition price remains subject to adjustments.

NVIDIA was already both an investor in Hugging Face and a major contributor to its platform. The chip company says it has published more than 500 models and more than 250 open datasets there. Hugging Face has simultaneously supported models and tools designed for hardware from NVIDIA’s competitors.

2. The companies promise an open, hardware-neutral platform

The central operational commitment concerns Hugging Face’s neutrality. NVIDIA says developers will continue to choose their own models, frameworks, cloud providers, inference services and computing platforms. Its announcement explicitly states that NVIDIA hardware will not be required to build on Hugging Face or deploy software through it.

The SEC filing gives this promise more specific boundaries. NVIDIA committed to allowing users and model makers to upload and download models and datasets of their choosing, consistent with Hugging Face’s existing practices. It also said the platform would continue supporting other silicon vendors.

NVIDIA separately promised continued support for both open-source and open-weight models from model developers across the ecosystem. It said Hugging Face would retain multi-cloud and multi-accelerator development and deployment, allowing customers to select hardware and infrastructure according to their own requirements.

These commitments address an immediate concern created by the transaction. Hugging Face’s usefulness depends partly on being a common venue for software that can run across competing hardware, cloud and inference environments. Ownership by one accelerator supplier creates an incentive conflict even if the public platform remains accessible.

The commitments are confirmed statements of NVIDIA’s current plans, but they are not evidence of future execution. The deal has not closed, and the company has not published detailed governance arrangements, technical enforcement mechanisms or contractual remedies for users if its platform policies change later.

Delangue went further in his announcement, saying Hugging Face would remain “open, independent and compute agnostic.” Corporate independence cannot retain its previous meaning once the company becomes part of NVIDIA. The more measurable commitments are therefore the ones covering model choice, uploads and downloads, competing accelerators and cloud-provider support.

3. Why the acquisition fits NVIDIA’s open-model strategy

NVIDIA benefits when developers train, fine-tune and run computationally demanding models, regardless of which organization originally created them. Open models can expand that activity by allowing companies, universities and individual developers to adapt existing weights rather than obtain every capability through a closed API.

Acquiring Hugging Face would give NVIDIA a direct position in the software distribution layer used by much of that community. It could fund additional storage, security, evaluation and deployment infrastructure while integrating its own models, development libraries and computing services more closely with the platform.

The commercial logic does not require NVIDIA to exclude competing hardware. A large, active Hugging Face ecosystem can increase overall demand for AI computation, including workloads that eventually run on NVIDIA products. Preserving broad participation may therefore be more valuable than converting the Hub into an exclusive storefront.

Independent reporting also indicates that Hugging Face approached NVIDIA and had received interest from other potential buyers. The transaction is consequently not described as an unsolicited takeover. Delangue said he initiated discussions with NVIDIA CEO Jensen Huang because scaling open-source AI required more compute, collaboration and support.

For developers, however, ownership still matters even without technical exclusivity. NVIDIA would control a platform that provides visibility into which models, architectures and tools attract attention across the ecosystem. That position could inform its software and hardware priorities. It also places decisions about platform policies, discovery and infrastructure investment inside a company that sells products to many of the same developers and model providers.

4. Closing conditions and regulatory exposure

The announcement is an agreement to acquire Hugging Face, not a completed acquisition. NVIDIA expects closing during the first half of 2027, and its SEC filing makes clear that regulatory approvals are required. Until the transaction closes, Hugging Face remains a separate company.

NVIDIA identified government restrictions on open-source AI as a transaction-related risk. Its filing says new rules could govern the development, training, release, distribution, transfer, deployment or use of AI models. Such measures could limit the models and datasets available through Hugging Face, require platform changes, increase compliance costs or produce investigations and enforcement actions.

The company specifically highlighted the international character of the open-model ecosystem. NVIDIA said many widely used open-source models originate in China before developers elsewhere download, revise, fine-tune and test them. Regional restrictions could therefore affect both Hugging Face’s catalog and NVIDIA’s ability to realize the expected benefits of the acquisition.

For existing Hugging Face users, no immediate migration or hardware change has been announced. NVIDIA’s stated policy is that developers will not need its compute to use the platform and will retain access to other accelerator and cloud options. The practical test will come after closing, through product integrations, pricing, model-distribution policies and the treatment of competing infrastructure providers.

Frequently Asked Questions

Has NVIDIA completed the Hugging Face acquisition?

No. The companies signed a definitive agreement on September 2, 2026, but expect the transaction to close in the first half of 2027 after required regulatory approvals and other closing conditions.

How much will NVIDIA pay?

NVIDIA announced a headline value of $12,930,300,000. Its SEC filing specifies approximately $11.9 billion for Hugging Face stockholders and up to approximately $1 billion in equity-based employee retention awards.

Will Hugging Face require NVIDIA GPUs?

NVIDIA says no. It has committed to retaining multi-accelerator support and states that NVIDIA compute will not be required to build on or deploy through Hugging Face.

Will models and datasets from other companies remain available?

NVIDIA says Hugging Face will continue supporting models and datasets from across the ecosystem, including uploads and downloads chosen by users and model makers. Future availability may still be affected by laws or regulatory restrictions.

Are Hugging Face’s founders leaving?

Delangue said the founders and team are staying to continue the company’s open-source AI mission. NVIDIA’s retention program provides up to approximately $1 billion in equity-based awards for employees joining the company.

Sources

• Clément Delangue’s original acquisition announcement on X
• NVIDIA: NVIDIA to Acquire Hugging Face
• NVIDIA Form 8-K filed with the US Securities and Exchange Commission
• Associated Press: Nvidia to spend $13 billion on Hugging Face
• Axios: Nvidia buying Hugging Face for nearly $13B